Ant Group's Inclusion AI has open-sourced the Ling-2.6-1T flagship model with 1 trillion total parameters, activating only 63B per inference. Under MIT license, the MoE model features a 'fast thinking' strategy that cuts token waste.
Ant Group's Inclusion AI open-sourced Ling-2.6-flash, a 104-billion-parameter MoE model that activates only 7.4B per token and achieves state-of-the-art results on agent benchmarks.
Ant Group's Inclusion AI has released Ling-2.6-1T, a trillion-parameter model that prioritizes fast thinking and token efficiency over deep reasoning. It's designed for agent workflows and is already available via a free API on OpenRouter.
Ant Group's AI lab has confirmed its new Ling-2.6-flash model is the previously anonymous 'Elephant Alpha' on OpenRouter. It's a 104B-parameter model built for efficient AI agents, promising significantly lower token usage and higher throughput than rivals.
